U.S. patent application number 10/102667 was filed with the patent office on 2003-09-25 for personal choice biometric signature.
Invention is credited to Chong, Delano P., Wong, Jacob Y..
Application Number | 20030179909 10/102667 |
Document ID | / |
Family ID | 36061678 |
Filed Date | 2003-09-25 |
United States Patent
Application |
20030179909 |
Kind Code |
A1 |
Wong, Jacob Y. ; et
al. |
September 25, 2003 |
Personal choice biometric signature
Abstract
A biometric method and system for personal authentication using
sequences of partial fingerprint signatures provides a high
security capability to various processes requiring positive
identification of individuals. This approach is further enhanced by
employing a frequency domain technique for calculating a Similarity
Index of the partial fingerprint signatures. In a baseline usage,
the sequential partial fingerprint sequence techniques augments
sentinel systems for gaining access to restricted areas, and when
used in combination with financial cards, offer a unique and
greatly simplified means for authenticating or identifying
individuals. A highly automated technique initially obtains four
(illustratively) linear partial fingerprint signatures which serve
as reference data against which later proffered candidate data in
the form of at least two linear partial fingerprint signatures are
compared for authentication. The particular two candidate
signatures used and the sequence in which they are submitted are
selected with the user's consent and serve as a PIN-like unique
personal code. In an advanced embodiment, the same two candidate
signatures in the chosen sequence are processed in a unique FFT/DFT
process to produce a highly reliable Similarity Index to
authenticate or verify the identity of individuals. The use of only
partial fingerprint data greatly allays the concerns of widespread
fingerprint dissemination by many individuals.
Inventors: |
Wong, Jacob Y.; (Goleta,
CA) ; Chong, Delano P.; (Mountain View, CA) |
Correspondence
Address: |
JAMES F. COTTONE
Suite 403
2001 Jefferson Davis Highway
Crystal Plaza One
Arlington
VA
22202
US
|
Family ID: |
36061678 |
Appl. No.: |
10/102667 |
Filed: |
March 22, 2002 |
Current U.S.
Class: |
382/115 |
Current CPC
Class: |
G06V 40/1335 20220101;
G07C 9/257 20200101; G06V 40/1347 20220101 |
Class at
Publication: |
382/115 |
International
Class: |
G06K 009/00 |
Claims
1. A biometric method of authenticating the identity of an
individual employing a preselected sequence of linear partial
fingerprint signatures processed in the frequency domain,
comprising the steps of: (a) obtaining a reference set of linear
partial fingerprint signatures from an individual and
electronically performing a Fourier transform on said reference set
to obtain a frequency domain reference set, (b) storing a first
electronic representation of said reference set in a comparison
means, said first electronic representation selected from a group
including said reference set and said frequency domain reference
set; (c) generating a candidate set of linear partial fingerprint
signatures from an individual and electronically performing a
Fourier transform on said candidate set to obtain a frequency
domain candidate set; (d) entering a second electronic
representation of said candidate set into said comparison means,
said second electronic representation selected from a group
including said candidate set and said frequency domain candidate
set; (e) providing a comparison means for comparing said first and
second electronic representations, (f) calculating in said
comparison means a dot product of said frequency domain reference
set with the complex conjugate of said frequency domain candidate
set, or alternately, a dot product of said frequency domain
candidate set with the complex conjugate of said frequency domain
reference set; and (g) formulating a similarity index from said dot
product calculation, said similarity index resulting in the
production of a real number between 0 and 1 resulting from said
formulating, whereby the identity of an individual providing said
reference and candidate sets may be authenticated if said
similarity index is greater than a predetermined level between 0
and 1.
2. The biometric method of claim 1 including the further step of
calculating in said comparison means the normalized FFT of said
frequency domain reference set and said frequency domain candidate
set, by dividing each set by the square root of the dot product of
itself with its complex conjugate to produce a normalized frequency
domain reference set and a normalized frequency domain candidate
set, and thereafter formulating said similarity index.
3. The biometric method of claim I wherein said generating step
further includes the process wherein said candidate set is derived
from at least two scans taken along predetermined paths on that
individual's fingerprint, the particular two of said at least two
scans and the sequence of their application to said comparison
means being selected so as to serve as a personal code.
4. The biometric method of claim 1 wherein performing said Fourier
transform steps are accomplished using an FFT process.
5. The biometric method of claim 4 wherein said FFT process is
carried out following a DFT process on at least one of said
reference and candidate sets.
6. The biometric method of claim 1 including the further step of
establishing a first authentication level and a second
authentication level within said comparison means for quantifying
the similarity index formulation result, whereby the identity of an
individual is authenticated if said real number is greater than
said first level and the identity of an individual is
non-authenticated if said real number is less than said second
level.
7. The biometric method of claim 6 wherein said first
authentication level is a number between 0.7 and 1 and said second
authentication level is a number between 0 and 0.3.
8. The biometric method of claim 6 further including providing an
output signal generator as part of said comparison means, for
providing an enabling signal to an external device upon
determination that the identity of an individual is authenticated
by said similarity index result number exceeding said first
level.
9. The biometric method of claim 1 wherein said generating step is
accomplished by carrying out a timed series of sample data points
to produce said said candidate set of linear partial fingerprint
signatures, said samples taken at uniformly spaced time intervals,
whereby the number of sample data points in said candidate set may
be adjusted in said comparison means to compensate for fingerprint
movement speed while generating said candidate set with respect to
said obtained reference set.
10. The biometric method of claim 1 wherein said obtaining step
includes producing a reference set of linear partial fingerprint
signatures having a number of sample data points provided at
uniformly spaced time intervals, whereby the number of sample data
points in said reference set may be adjusted in said comparison
means to compensate for fingerprint movement speed while obtaining
said reference set with respect to said generated candidate set
11. A biometric method of authenticating the identity of an
individual employing a preselected sequence of linear partial
fingerprint signatures comprising the steps of: (a) obtaining a
reference set of linear partial fingerprint signatures from an
individual and storing an electronic representation of the
reference set in a comparison means; (b) generating a candidate set
of linear partial fingerprint signatures from an individual and
entering an electronic representation of the candidate set into
said comparison means, said candidate set derived from at least two
scans taken across corresponding predetermined paths on that
individual's fingerprint, the particular two of said at least two
scans and the sequence of their entering into said comparison means
being selected so as to serve as a personal code; (c) providing a
comparison means for comparing said reference and candidate sets
and for providing an affirmative response for a successful
comparison and a negative response for an unsuccessful comparison;
and (d) wherein said reference set is derived from not more than
four bidirectional scans.
12. The biometric method of claim 11 wherein said not more than
four bidirectional scans are taken along two straight lines
intersecting at their midpoints, each line rotationally displaced
90 degrees from the other.
13. The biometric method of claim 12 wherein said reference set
includes four linear partial fingerprint signatures and said
candidate set includes two out of a possible sixteen linear partial
fingerprint signatures.
14. The biometric method of claim 11 wherein said comparing means
is included in a microprocessor embedded within an
identification/credit card.
15. The biometric method of claim 11 wherein said affirmative
response is one or more responses selected from a group including
an alphanumeric visual indication, a color-coded visual indication
and a signal for controlling an external control unit.
16. A biometric system for authenticating the identity of an
individual employing a preselected sequence of linear partial
fingerprint signatures processed in the frequency domain comprising
(a) means for obtaining a reference set of linear partial
fingerprint signatures from an individual, said reference set
derived from at least one scan taken across said individual's
finger and electronically performing a Fourier transform on said
reference set to obtain a frequency domain reference set; (b) means
for storing a first electronic representation of said reference set
in a comparison means, said first electronic representation
selected from a group including said reference set and said
frequency domain reference set; (c) means for generating a
candidate set of linear partial fingerprint signatures from an
individual, said candidate set derived from at least two scans
taken across predetermined bidirectional portions of said
individual's finger and electronically performing a Fourier
transform on said candidate set to obtain a frequency domain
candidate set, (d) means for entering a second electronic
representation of said candidate set into said comparison means,
said second electronic representation selected from a group
including said candidate set and said frequency domain candidate
set; (e) comparison means for comparing said first and second
electronic representations, (f) means for calculating in said
comparison means a dot product of said frequency domain reference
set with the complex conjugate of said frequency domain candidate
set, or alternately, a dot product of said frequency domain
candidate set with the complex conjugate of said frequency domain
reference set, and (g) means for formulating a similarity index
from said dot product calculation, said similarity index resulting
in the production of a real number between 0 and 1 resulting from
said formulating, whereby the identity of an individual providing
said reference and candidate sets may be authenticated if said
similarity index is greater than a predetermined level between 0
and 1
17. The biometric system of claim 16 wherein said means for
obtaining includes an optical scanner for deriving said at least
one scan while said individual's finger is held stationary.
18. The biometric system of claim 16 wherein said means for
generating includes a single element optical sensing device for
deriving said at least two scans while said individual's finger is
moving relative to said single element optical sensing device.
19. The biometric system of claim 16 wherein said comparison means
includes a microprocessor and said means for storing comprises a
RAM/ROM portion of said microprocessor.
20. The biometric system of claim 19 wherein said microprocessor is
embedded within an identification/credit card having a
self-contained power source.
21. The biometric system of claim 20 wherein said microprocessor
further includes speed compensating means for assuring that said
means for obtaining and said means for generating produce sets of
linear partial fingerprint signatures that temporally facilitate
said similarity index formulating
22. The biometric system of claim 16 wherein said comparison means
further comprises an output circuit for providing control signals
to an external control unit, said output circuit employing wired or
wireless channels for said controlling.
23. The biometric system of claim 16 wherein said comparison means
further comprises a visual indicator for displaying an output from
said means for formulating to denote a condition where an
individual's identity is authenticated.
24. The biometric system of claim 16 wherein said comparison means
further comprises a scanning device having a template with two sets
of contoured and indented scanning tracks and a single optical
element sensor unit located at the center of said template
25. The biometric system of claim 24 wherein said scanning device
serves as said means for generating.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] The present application is a Continuation-in-Part of a U.S.
Non-Provisional Patent Application filed on Feb. 13, 2002 for
"Authentication Method Utilizing A Sequence of Linear Partial
Fingerprint Signatures Selected By A Personal Code," which is
incorporated by reference in full herein.
TECHNICAL FIELD
[0002] The present invention relates generally to the field of
personal authentication. In particular, this invention pertains to
a comparison method of utilizing a specific sequence of an
individual's linear partial fingerprint signatures selected by a
Personal Code as a basis of comparison for the authentication or
identification of the individual in question.
BACKGROUND
[0003] In the complex society that we are living in, there are
numerous occasions where individuals have to authenticate
themselves by means other than personal recognition. Until
recently, a common approach to this has been the issuance of
personal identification cards which range in complexity depending
on the purpose for which they are to be used. For situations that
are deemed only of secondary importance, the cards may merely
contain the individual's name, signature and an identification
number. Here, the presentation of the card will be proof enough of
the user's identity if the card signature matches that of the
user's as taken at the time of use. For situations that require a
more positive identification, such cards are also provided with the
individual's photograph, as in the case for driver's licenses and
passports.
[0004] Unfortunately, these identification instruments have become
the common victims of illegal falsification and duplication. The
rampant credit card fraud of recent years has certainly accentuated
the inadequacy of using such personal instruments to authenticate
oneself in many instances. To this end when bank-issued ATM cards
were finally accepted and used by the American public in large
numbers in the middle 1980's, a new identification means was
introduced in the form of what is now called a PIN number, or
Personal Identification Number, which typically takes the form of
an easily-memorisable 4-digit decimal number.
[0005] Even though there exist superior ways and methods for use in
identifying or authenticating an individual, particularly those
that use one's natural body codes such as faces, fingerprints,
retina patterns, irises and voice prints, they have only been
deployed to date in highly special circumstances where the absolute
security of one's identity warrants the additional complexity.
Indeed the use of fingerprints to identify unique individuals has
been around for well over a hundred years. Either "rolled"
fingerprint or "flatly placed" fingerprint inked impressions are
commonly used and the identification can be classified as "passive"
because the individual is not required to perform any finger
motions during the subsequent process of identification. As is
well-known, in collecting the so-called "rolled" fingerprint
impressions, an individual's inked thumb or other fingers is
rotated from one side of the nail to the other so that the entire
pattern area can be printed on paper. Characteristic features or
patterns of fingerprints such as "arches", "loops" and "whorls"
(referred to as keys) are routinely employed by
fingerprint-identifying technicians to define fingerprint patterns
for easier comparison and identification of them. The so-called
Henry classification system is often used to determine if two
prints are the same even though this system requires a skilled
expert to compare the individual characteristics of the prints.
[0006] The classical approach of using fingerprints to identify
individuals, albeit among one of the best known to date, is
nevertheless rather complex and may require elaborate optical
instruments such as high-power microscopes for detailed fingerprint
pattern examinations. Collection of inked fingerprint impressions
can be rather messy and also takes operator skill and a finite
amount of time in order to do an adequate job. As mentioned above,
identification of fingerprints belonging to unique individuals
using comparison methods requires trained experts or experienced
technicians. Furthermore identification of individuals via
fingerprint matching is not really an exact science and is by no
means 100% objective or accurate. Added to all these is the fact
that an individual's fingerprints are not safe or fully protected
from fraudulent use because most people frequently and
inadvertently leave behind fingerprints while performing their
daily routines. These fingerprints can be willfully recovered for
illegal use as falsified personal identifications.
[0007] Not surprisingly, not all people feel comfortable in
submitting their fingerprints for their personal identification
such as credit cards, employee entrance cards in workplaces etc.
except for very serious matters such as extreme security check for
sensitive federal appointments or for crime solving. One important
reason behind this is the fact that there is an undesirable stigma
of "criminal nature" associated with the use of fingerprints as a
method of identification. Replacing specially trained and
experienced fingerprint-identifying technicians requires the use of
very complicated detection machines equipped with complex
processing algorithms. These equipment are therefore necessarily
expensive. Still, in an effort to try to thwart the rampant credit
card fraud, proposals have been advanced over the past several
years to utilize one's fingerprint as a more secure way of
authenticating credit card holders. The use of fingerprints along
with the use of the so-called "smart cards", namely cards that
encapsulate a secure smart integrated circuit (IC) chip in the
plastic in lieu of the fraud-prone magnetic stripe for storing
sensitive and personal financial data, would surely eliminate once
and for all the credit card fraud problem existing today. The
development of the so-called biometric smart card using fingerprint
template identification has been on-going for a number of years but
unfortunately is still far from being a reality because of credit
card size and the cost constraints of this method, in addition to
having to overcome very difficult technical challenges.
[0008] However, the technical obstacles that have been encountered
to date in the implementation of the full-blown fingerprint
identification approach in the biometric smart card should not be
the determining factor in deciding whether or not this venerable
identification method should be deployed in the future.
Furthermore, the deployment of the retina pattern, iris and the
voice print as better and alternate ways to identify individuals
will likely encounter the same constraint problems in size, cost
and technical challenge without the benefit of a head start like
the use of fingerprints. Today the rampant credit card fraud
problem has not gone away. As a matter of fact, the problem grows
worse and more serious everyday that passes. Thus there presently
exist ample reasons why a new and better methodology is needed in
order to exploit the use of fingerprints as a secure way of
authenticating individuals, especially in circumstances of primary
importance like access to restricted area or restricted
information, or authorization of credit cards, without the existing
encumbrances of using fingerprints for identification as discussed
earlier above.
[0009] Ample prior art can be found in fingerprint detection
apparatus and methodology of using fingerprints for personal
authentication and identification. A list of earlier issued U.S.
patents relating to the prior art has been presented in a U.S.
application filed Feb. 13, 2002 for "Authentication Method
Utilizing a Sequence of Linear Partial Fingerprint Signatures
Selected by a Personal Code", of which the present application is a
continuation-in-part. Additional prior art dealing with
two-dimensional fingerprint images, their acquisition methodology
and apparatus, and their classification, interpretation and
comparison are presented as follows.
[0010] In U.S. Pat. No. 5,933,515 issued to Pu et al. in 1999, an
identification system using biometric information of human body
parts and a secret sequence code was advanced. In particular,
biometric information of human body parts is used to form the
secret sequence code. Specifically, a combination entry device
recognizes user's fingerprints which are entered as a sequence. The
fingerprints must be entered in the proper sequence in order to be
recognized by the system. Although in principle this invention has
a lot of merit and was among the first to introduce the concept of
giving an individual a choice in how to use his biometric
information as a way of his identification, only the use of
different complete fingerprints to form the biometric sequence was
taught. Thus the implementation of this teaching is in reality
extremely cumbersome and time consuming and it is certainly not
amenable to simple and low-cost realization in order to be
universally practical.
[0011] In U.S. Pat. No. 5,982,913 issued to Brumbley et al. in
1999, a method of fingerprint verification was advanced that
includes the steps of capturing a complete fingerprint of a number
of enrollees; capturing a portion of a claimant's fingerprint,
where the portion is less than an entire fingerprint; dividing the
portion of the claimant's fingerprint into a number of segments;
comparing each of the segments against the fingerprint of the
enrollee the claimant claims to be; generating a correlation score
for each of the segments; calculating a distance error for the
segments; combining the distance errors into an average distance
error; generating a verification vector based on each of the
correlation scores for each of the segments and the distance error;
establishing a threshold vector; and comparing the verification
vector against the threshold vector in order to determine whether
or not the claimant is the enrollee the claimant claims to be. It
is clear that the primary objective of the inventors is to simplify
the fingerprint verification process by devising means for
comparing portions of a complete fingerprint with a complete
reference fingerprint. The means used to achieve this objective are
still overly complicated and are not easily amenable to simple and
low-cost implementation.
[0012] In U.S. Pat. No. 6,226,391 issued to Dydyk et al. in 2001, a
method and apparatus for automatically placing a first unknown
image, such as an unknown fingerprint image, into one of a
plurality of categories. The invention includes storing in a
library a plurality of value series, each of which series is
derived from the frequency representation of an image category. The
categorization process and apparatus takes the frequency image of a
first unknown pattern to create a first frequency image. The
frequency image plane of the first (unknown) frequency image is
divided into a plurality of frequency image plane regions. Each of
the frequency image plane regions may be an angular segment
radiating from the origin of the frequency image plane. A region
value is assigned to each of the frequency image plane regions
based on the total energy in the frequency image in that region.
The region values for the first frequency image are combined to
generate a first series of region values. The first series of
region values is compared in a comparator with each of the stored
value series. The comparator preferably performs a correlation
function on the pattern or series of the regional values using the
one dimensional frequency transform of the spatial representation
of the pattern or series of regional values.
[0013] Although this invention discloses the concept of fingerprint
classification using spatial frequency representation, correlation
functions and one-dimensional frequency transform, the apparatus
advanced to generate the frequency image of an unknown pattern in
order to utilize such a classification methodology is rather
complex and is certainly not amenable to simple and low-cost
implementation.
[0014] In U.S. Pat. No. 6,241,288 issued to Bergenek et al. in
2001, a novel fingerprint identification/verification system was
disclosed. This system uses bitmaps of a stored fingerprint to
correlate with a bit map of an input fingerprint, wherein an
accurate reference point is located. This is followed by the
selection of several two-dimensional areas in the vicinity of the
reference point of the input image of the fingerprint. These areas
are then correlated with stored fingerprint recognition information
to determine if the input fingerprint image and the stored
fingerprint recognition information are sufficiently similar to
identify/verify the input fingerprint. It can be seen from this
brief summary of the patent, the teaching of this invention is very
complex and unlikely to be able to be implemented simply and in a
low-cost manner.
[0015] Additional teachings of fingerprint identification systems
and methods of related interest, particularly in the use of Fast
Fourier Transform techniques for 2-dimensional fingerprint image
analysis may be found in other U.S. patents, including--U.S. Pat.
No. 5,910,999 issued to Mukohzaka in 1999; U.S Pat. No. 59,915,034
issued to Nakajima et al. in 1999; U.S. Pat. No. 5,999,637 issued
to Toyoda et al. in 1999; U.S. Pat. No. 6,024,287 issued to Takai
et al. in 2000; U.S. Pat. No. 6,075,876 issued to Draganoff in
2000; U.S. Pat. No. 6,094,499 issued to Nakajima et al. in 2000;
U.S. Pat. No. 6,341,028 issued to Bahuguna et al. in 2002; and U.S.
Pat. No. 2002/0018585 Al issued to Kim in 2002.
[0016] There is hardly any doubt that the prior inventions
summarized above and those presented in the aforementioned U. S.
application filed on February 13, 2002, of which the present
application is a continuation-in-part, have made significant
progress towards simplifying the overall mechanics for the
acquisition, classification and comparison of fingerprints. They
have also removed in some cases the subjectivity and ambiguity in
the employment of the well-known Henry classification system to
determine if two prints are the same. However, the conventional
thinking of using the entirety or even portions of one's complete
fingerprint on a comparison basis with stored counterparts to
authenticate oneself is still today far too complex a task to
accomplish simply and economically, despite the availability of
clever correlation methods and high-power mathematical tools.
[0017] The approach taken by the aforementioned U.S. application
and the present continuation-in-part represent major departures
from such thinking. Certain specific and well-defined partial
fingerprints (e.g. linear or straight line segments) belonging to
an individual are now looked upon as his biometric signatures and
he has a choice of using linear signatures in a self-chosen
sequence for his unique authentication. Unlike one's written
signature in the past which was used for his identification but
suffered from frequent illegal falsification and duplication, today
one's written signature can be replaced by a so-called Personal
Choice Biometric Signature (PCBS) or sequential linear signatures
by choice which can neither be duplicated nor falsified for fraud.
The reason for advancing this new kind of thinking is that
comparing an individual's entire fingerprint against one's stored
complete fingerprint is not cost effective in low-cost applications
including smart card systems primarily because of the size and
expense of the scanner required to capture an entire fingerprint
and partly because of the amount of memory required to store and
process one's entire fingerprint.
OBJECTS AND SUMMARY OF THE INVENTION
[0018] It is therefore a primary object of the present invention to
provide methods and apparatus for verifying or authenticating the
identity of individuals using only partial fingerprint data.
[0019] A further object of the present invention is to provide
methods and apparatus for verifying or authenticating the identity
of individuals using frequency domain techniques applied to partial
fingerprint data.
[0020] A yet further object of the present invention is to provide
improved high security methods and systems for personal
authentication or identification using sequences of partial
fingerprints (linear signatures) selected by a personal code.
[0021] A still further object of the present invention is to
provide a method of verifying or authenticating the identity of
individuals using a sequence of at least two partial fingerprints
(linear signatures) taken from a group of more than two linear
signatures where the particular two used, as well as the particular
sequence in which they are used, is arranged to constitute a
PIN-like user's personal code.
[0022] A still further object of the present invention is to
provide a system comprising a Selective Partial-fingerprint
Authenticator (SPA) and accompanying novel software algorithm for
first obtaining and processing a "reference" set of partial
fingerprints (linear signatures) from an individual and
subsequently a "candidate" set of at least two partial fingerprints
(linear signatures) from that individual in order to authenticate
the individual's identity.
[0023] Additional objects of the present invention are: to advance
a method for simplifying the use of one's fingerprint to
authenticate one's identity via the Personal Choice Biometric
Signature (PCBS) analogous to one's written but fraud-prone
signature of the past; to facilitate the replacement of specially
trained and experienced fingerprint identifying technicians with
relative simple, ultra-small-sized and low cost device that can be
manufactured in high volumes and thereby renders the task of
fingerprint authentication simpler, less costly and less subjective
to personal opinion; to safeguard the use of fingerprints to
identify individuals from the illegal recovery of fingerprints and
their subsequent fraudulent use.
[0024] Although the present invention still uses one's basic
fingerprint as a means to authenticate one's identity
(authentication and identification are hereinafter used
substantially interchangeably), it departs significantly from the
manner of its traditional utilization. Instead of using the entire
inked impression of fingerprints ("rolled" or "flatly placed") and
its associated characteristic features or patterns such as
"arches", "loops" and "whorls" for identification purposes, only
certain pre-defined partial fingerprints are used and they are
designated herein as "linear signatures". They are so named because
these linear signatures represent different linear (straight line
segment) image scans of the ridges and troughs of a fingerprint all
through its reference center. Two or more of these linear
signatures, augmented by the incorporation of one's own choice in
selecting which linear signatures and their respective application
sequence, are used to authenticate an individual. This is likened
to remembering a PIN number but instead of punching in the
traditional 4-digit PIN, one replaces that with two or more simple
and sequential "strokes" of one's index finger on a well-marked and
contoured template. It is the choice of making any two or more out
of many possible "strokes" and their respective sequence that in
essence replaces the use of the PIN. We have in essence replaced
the comparison of one's entire fingerprint with one's a priori
stored counterpart for authentication with a much simpler procedure
using the Personal Choice Biometric Signature (PCBS) discussed
earlier for one's identification.
[0025] Unlike the prior art presented earlier, these linear
signatures referred to above are not captured via the use of a
linear or matrix array of sensors which are costly and whose
numerous outputs are rather complex to process. Instead, only a
single sensor located at the reference center of a well-marked and
index-finger contoured template, which forms part of the Selective
Partial-fingerprint Authenticator (SPA), is used. This template, an
integral part of an SPA, is used to capture the a priori reference
linear signatures of an individual according to his/her personally
selected sequence (personal code) and that will constitute as
his/her PCBS. At subsequent times one's PCBS, captured using the
same or a structurally equivalent template portion of an SPA, will
be used to uniquely authenticate the individual. The use of the
well-marked and index-finger contoured template for both instances
is to minimize the spatial variation of a particular straight line
segment of one's fingerprint (linear signature) as seen by the
single sensor of the SPA when the index finger moves over it. In
addition, in order to simplify the capture of the PCBS, the
template is designed to allow only two orthogonally oriented
"strokes" of the index finger, arranged like the arithmetic "+"
symbol. As will be explained in more detail below, there are only
two grooves on the top surface of the template oriented 90 degrees
to each other with the single sensor located at their
intersection.
[0026] A novel algorithm formulated around the use of Discrete
Fourier Transform (DFT) and Fast Fourier Transform (FFT) techniques
as applied to the captured linear signatures is used to generate a
unique Similarity Index (SI) from one's reference PCBS and one's
subsequently submitted candidate PCBS. Based upon the value of the
calculated SI, one's positive or negative authentication can be
readily and quantitatively determined utilizing the reference and
submitted candidate PCBS's of the individual.
[0027] Thus the present invention, in addition to advancing the
method of using only one's selected partial fingerprints for
authentication, thereby giving the individual a PIN-like
protection, also simplifies the procurement of the so-called linear
signatures and the PCBS with the use of only one sensor located
strategically at the reference center of a specially designed
template which is part of the Selective Partial-fingerprint
Authenticator (SPA). Furthermore a novel parameter called the
Similarity Index (SI) is formulated and introduced applying DFT and
FFT techniques on the linear signatures. This Similarity Index (SI)
can be easily and quantitatively used to discern accurate
similarity or dissimilarity from individual's submitted reference
and candidate PCBS's for one's authentication.
BRIEF DESCRIPTION OF THE DRAWINGS
[0028] FIG. 1 depicts four linear signatures and their respective
scanning direction, as superimposed over a typical fingerprint
central portion;
[0029] FIG. 2 depicts a simplified identification template device
with the four scanning directions well delineated, and a sensor
unit at the center of the template;
[0030] FIG. 3 depicts the scanning motion for a finger of an
individual having a Personal Code {13}
[0031] FIG. 4 is a simplified block diagram for the Selective
Partial-fingerprint Authenticator (SPA) according to the present
invention;
[0032] FIG. 5 schematically shows details of a fingerprint sensor
unit;
[0033] FIG. 6 depicts the definition of the fingerprint reference
center of the bottom side of an individual's finger.
[0034] FIG. 7 shows a typical fingerprint ridge and valley pattern
in simplified form along a scanned direction as illustrated;
[0035] FIG. 8 depicts the linear signature in analog form, as
sensed by a fingerprint sensor unit of FIG. 5 corresponding to the
scanned direction of FIG. 7;
[0036] FIG. 9A depicts a complete linear signature of an individual
as generated by an SPA device;
[0037] FIG. 9B depicts the linear signature of FIG. 9A after
selected data from both edges of the finger have been deleted; FIG.
9C shows the Fast Fourier Transform (FFT) of the linear signature
shown in FIG. 9B;
[0038] FIG. 10 is a flow chart showing the formulation steps for
the Similarity Index (SI) method according to the present
invention;
[0039] FIG. 11A shows a reverse linear signature (i.e. scanning in
the opposite direction) of that shown in FIG. 9B;
[0040] FIG. 11B shows the Fast Fourier Transform (FFT) of the
linear signature shown in FIG. 10A;
[0041] FIG. 12A shows the slightly modified linear signature of
FIG. 9B by reducing the amplitude of some of the peaks and
valleys;
[0042] FIG. 12B shows the Fast Fourier Transform (FFT) of the
linear signature shown in FIG. 12A;
[0043] 13A shows a new and different linear signature from that
shown in FIG. 9B;
[0044] FIG. 13B shows the Fast Fourier Transform (FFT) of the
linear signature shown in FIG. 13A,
[0045] FIG. 14A shows a complete and non-truncated linear signature
of that shown in FIG. 9B;
[0046] FIG. 14B shows the linear signature as shown in FIG. 4A with
.about.15% of the data truncated from its beginning;
[0047] FIG. 14C shows the linear signature as shown in FIG. 14A
with .about.15% of the data truncated from its end, FIG. 14D shows
the linear signature as shown in FIG. 14A with .about.15% of the
data symmetrically truncated from its beginning and end
[0048] 15A shows the linear signature of that shown in FIG. 9B
taken with a larger aperture of the SPA device,
[0049] FIG. 15B shows the reverse linear signature (i.e. scanning
in the opposite direction) of that shown in FIG. 15A,
[0050] FIG. 15C shows the linear signature of that shown in FIG.
13A taken with a larger aperture of the SPA device;
[0051] FIG. 15D shows the reverse linear signature (i.e. scanning
in the opposite direction) of that shown in FIG. 15C;
[0052] FIG. 16A is a reference linear signature of an individual
taken with some nominal and acceptable finger movement speed;
[0053] FIG. 16B is a candidate linear signature of the same
individual as of FIG. 16A taken with half the nominal finger
movement speed; and
[0054] FIG. 16C is a candidate linear signature of the same
individual as of FIG. 16A taken with twice the nominal finger
movement speed.
BEST MODE FOR CARRYING OUT THE INVENTION
[0055] Instead of using the entire inked impression of fingerprints
and their associated characteristic keys for identification
purposes, a preferred embodiment of the current invention employs a
total of four straight-line segments (linear signatures) all
crossing the nominal "reference" center of the fingerprint being
processed. Referring to FIG. 1 there is shown a typical fingerprint
central portion 10 over which are superimposed two straight lines
which intersect at their midpoints to produce what are defined as
four linear signature paths.
[0056] A horizontal line 1-2 running form West to East shows the
path for taking a linear signature designated as 1,2 as depicted by
arrows 1,2 (e.g. the path which originates at the western terminus
1 of the line 1-2 and ends at its eastern terminus 2). The same
horizontal line 1-2 shows the path for taking a linear signature
designated as 2,1 as depicted by arrow 2,1. Clearly, this second
linear signature 2,1 will produce the same fingerprint data as the
previous linear signature 1,2 but in reverse order. Similarly, a
vertical line 3-4 running form North to South shows the path for
taking a linear signature designated as 3,4, as depicted by the
arrow 3,4 (e.g. again the path which originates at the northern
terminus 3 of the line 3-4 and ends at its southern terminus 4),
and, the same vertical line 3-4 shows the path for taking a linear
signature designated 4,3 as depicted by the arrow 4,3. Also as
before, this linear signature 4,3 will produce the same fingerprint
data as the previous linear signature 3,4 but in reverse order.
[0057] Alternately stated, the two (illustratively) lines 1-2 and
3-4, when displaced 90 degrees (illustratively) apart and arranged
to centrally intersect give rise to four possible orientations or
linear signature designations. When any two of these four are
selected in a particular sequence (by the individual being
authenticated, for example), a total of 16 possible combinations
are available resulting in a unique PIN-like capability. This is
accomplished even while using only partial fingerprint data, which
in itself is an additional security level capability. A much higher
security level can be achieved by selecting any N of the four
possible orientations or linear signature designations in a
particular sequence. The total possible combinations available will
then equal to 4 to the Nth. For example, when the value of N is 4,
a total of 256 possible combinations are available. When the value
of N is 6, a total of 4,096 possible combinations are available.
Thus if one selects N equal to 6, the level of security is almost
equivalent to (within a factor .about.2) that for a 4-digit
PIN.
[0058] Referring now to FIG. 2 there is shown a simplified
representation of an identification template device 14 showing the
four linear signatures and their scanning directions, as well as a
centrally located single sensor element 16 at the template center.
The present invention encompasses the storage of the four linear
signatures for an individual's fingerprint (thumb or any finger) as
uniquely belonging to an individual. In a preferred embodiment, it
is contemplated that the index finger be used and that a
right-handed person uses his right index finger and a left-handed
person uses his left index finger. However, in order to uniquely
identify that individual, one only uses N out of the 4 to the Nth
possible combinations of N linear signatures in sequence as
selected by that individual (and/or and associated entity) for
their identification or authentication. For example, if one selects
a value for N equal to 2, then these two so-selected sequential
linear signatures by the individual are referred to as the
individual's Personal Code.
[0059] In order to represent an individual's Personal Code in a
more user-friendly manner, simply as a two digit number (here we
set N=2 for clarity of explanation), one translates the linear
signatures as previously described with the assistance of and
reference to FIG. 3. Thus if one's Personal Code is selected as 1,2
and 3,4 (e.g. first West to East, then North to South) the
two-digit code will simply be {13} or thirteen. And if one's
Personal Code is selected as 1,2 and 1,2 (e.g. West to East twice),
then the two digit code will be {11} or eleven, and so on. Each
single digit in bracket corresponds to what was described above as
the starting terminus of the straight line, and the two digits in
brackets represents the unambiguous and simplified designation
sequence of the two number Personal Code.
[0060] The authentication methodology of the current invention
calls for an individual to first indicate his Personal Code and the
subsequent authentication of that individual is based only upon the
use and comparison of these two selected sequence linear signatures
with the corresponding sequential two that had been previously
stored. For clarity, the authentication process is illustrated in
FIG. 3 for a Personal Code {13}, thirteen. Note that the first of
the two sequential scans {1} is from West to East, the second scan
{3} is from North to South. Those contoured indentations or
grooves, shown in dashed lines surrounding each track or path, help
guide finger movement to minimize spatial variations from scan to
scan. Note also that the finger must always point upwards (to the
North) as shown. The sixteen two-digit Personal Codes (for N=2)
available are, using the bracket scheme described above: {11-14},
{21-24}, {31-34} and {41-44}.
[0061] The current invention is in essence a two step process. Step
one involves the generation and storage of a set of four non-inked
partial fingerprint data (e.g. N=2) for an individual in the format
of four linear signatures 1,2; 2,1; 3,4 and 4,3 previously
designated and described, or their simplified equivalents {1 }
through {4} respectively, as shown in FIG. 3. These are considered
as the reference linear signatures. This step may be likened to
obtaining an individual's a priori inked fingerprint impressions in
the traditional sense but without the use of ink, and of course,
using only partial fingerprint data.
[0062] Step two involves the actual taking and processing of two
sequential non-inked linear signatures (e.g. for N=2) of an
individual according to his supplied Personal Code (these are
considered as the candidate linear signatures), for comparison with
the corresponding two sequential linear signatures that are stored,
at the time of the authentication process. Step two will typically
be carried out at a point-of-sale terminal; ATM, or other
authentication venue. The authentication is affirmative if there is
deemed sufficient match between the two sets of sequential linear
signatures. Otherwise the authentication is negative, and the
proposed transaction is denied.
[0063] It is evident from the description above that step one of
the two processes entailed in the current invention can be afforded
a rather sophisticated measurement setup without too much concern
about cost and size constraints. This measurement equipment could
be designed to scan linearly (in a substantially straight line)
with known spatial scanning speed in two different directions all
passing through a well-defined fingerprint "reference" center, the
lines each separated by an angle equal to 90 degrees. Since two of
the four linear signatures are basically the same data except for
scanning in the opposite directions, only two such directional
scans suffice to generate the four linear signatures with the
individual's index finger held stationary and its center coincident
with the reference center of the scanning surface (template) of the
equipment. The scanning speed information is important for the
subsequent pro-rating in time of the relevant stored linear
signatures in order to match those generated by the individual at
the time of authentication or identification. Such a pro-rating of
the scanning speed in time by the appropriate software installed in
the processor of the hardware for generating the linear signatures
of the individual at the time of authentication eliminates the
finger motion speed dependence on the generated linear signatures
as will be explained in more detail below.
[0064] Step two of the process for implementing the present
invention is accomplished by a Selective Partial-fingerprint
Authenticator SPA device that will procure or generate two (for
N=2) or more (for N>2) candidate linear signatures of the
individual corresponding to his entered Personal Code at the time
of authentication. The SPA device also holds the four reference
linear signatures and other relevant information of the individual
to be authenticated. Furthermore the SPA holds a microprocessor
installed with appropriate software for processing the generated
authentication information (here 15 the two or more linear
signatures) during an authentication process in order to generate
an acceptance signal based upon the result of a sufficient positive
or insufficient or negative match. In principle such an SPA device
can take many forms with varied complexity dependent upon the
purpose or application for which it is to serve. In any event this
device cannot afford the type of sophistication and luxury
installed in the measurement equipment used in step one of the
process where only a few equipment units would suffice to generate
all the reference linear signatures needed for use with this
invention.
[0065] One of the ideal applications for the current invention is a
security device employed along with an ordinary non-secure card to
simply identify an individual (name and ID number of an individual
only) for security purposes to control access to restricted area or
restricted information. One example is to authenticate workers
entering their workplaces. Here using the non-secure identification
card (either in the form of a magnetic card or smart card) the
worker enunciates who he or she is and then has to provide his
Personal Choice Biometric Signature (PCBS) or linear signatures
according to his Personal Code submitted earlier to his employer in
order to get access to the workplace. Another more relevant and
opportune application is to authenticate workers at airports
getting access to restricted areas in view of the country's
heightened security concerns.
[0066] For these specific applications, the SPA device suitable for
use in step two of the two-step process is shown in block diagram
form in FIG. 4. The blocks of an SPA device as depicted include a
scanning device 22, a sensor driver/signal preprocessor circuit 24,
a microprocessor 26 complete with its CPU, ROM, RAM and I/O units,
a magnetic reader head 28 with the slider track 30, a smart card
contact receptor 32 or an RF receiver (optional and not shown), an
LCD indicator 34, power supply circuit 36 with external AC cord 38
and a battery 40 for standby power in the event of a power failure.
An output circuit 27 driven from the microprocessor 26 provides
signals to an external control unit (not shown), which output
circuit 27 may employ wired or wireless (e.g. via RF) channels in
order to perform a specific external function such as the
unlatching of a lock in addition to the "positive authentication"
indication by the LCD indicator 34.
[0067] With the exception of scanning device 22 and the sensor
driver/signal preprocessor unit 24, the rest of the building blocks
in the SPA device 20 are straightforward and known to those skilled
in the art pertaining to the current invention Therefore, the
details of operational interactions between the microprocessor 26,
its output display LCD 34, its smart card input receptacle 32 and
power sources 36 and 40 etc. are not described further in the
interests of brevity, other than to note that the processed
representation of the linear partial fingerprint sequences (both
reference and candidate) may be stored in the ROM and/or RAM
portions of the microprocessor 26. The scanning device 22 further
comprises a template 40 with two sets of contoured and indented
scanning tracks, of which track 42 are typical, and a sensor unit
44 located at the center of the template 40. The design details for
the template 40 and its associated scan tracks 42 correspond to the
features previously depicted in connection with the description of
FIGS. 2 and 3. Thus the sensor unit 44 is located at the
fingerprint center as previously shown. The two sets of scanning
tracks 42 correspond to the scanning directions 1,2 and 3,4 of FIG.
2. The dashed lines surrounding the tracks 42 additionally indicate
that the tracks have an operational width set so as to alleviate
finger alignment problems during the authentication as well as to
mitigate software requirements.
[0068] The structural and functional details of the sensor unit 44
are described with reference now to FIG. 5. The sensor unit 44
comprises a special fixture (slab) 46 with a very small aperture 48
on its top surface 50 and a conical cavity 52 with an optically
opaque surface 54. The conical cavity 52 opens up at its bottom 56
to accommodate a small header 58 (e.g. TO-18) equipped with a
specially designed header can 60 which serves to hermetically seal
the devices (LED and silicon photodiode) die-attached onto the top
surface of the header 58. The header can 60 is equipped with a thin
transparent window 62 made out of quartz or sapphire for optical
radiation to pass into and out of the space 64 formed between the
top surface of the header 58 and the bottom side of the header can
60. The header can 60 is further equipped with an small aperture
tube 66 on its top so as to provide an aperture stop 68 for the
radiation emitted by a light-emitting-diode (LED) 70. Die-attached
onto the top of the header 58 are an LED 70 and a ring or
donut-shaped silicon photodiode 72 encircling the LED 70. The
electrical leads 74 of the silicon photodiode 72 are connected to
the signal preprocessor portion of the circuit 24 (of FIG. 4) and
the electrical leads 76 of the LED 70 are connected to the sensor
driver portion of circuit 24. Also shown in FIG. 5 is the
previously described template 40 being butted against by the top 50
of the slab 46 with the bottom side of an individual's finger 78
showing the ridges and troughs, of which the two shown as 80 are
typical, in touch with the other side of the template 40.
[0069] The sensor unit 44 is used to procure linear signatures for
the individual when his finger (bottom side down) moves along any
one of the four possible scanning tracks 42 at the time of
authentication. The so-called fingerprint reference center (12 of
FIG. 1) is ideally placed in the center of the contoured and
indented track as the finger moves along the desired track. With
brief reference to FIG. 6, the so-called fingerprint reference
center is defined vertically (from North to South) as the midpoint
between an apex 82 and the first break 84 of the prints and
laterally (East to West) as the midpoint of the finger 78.
[0070] As the finger 78 moves across the contoured and indented
track as was described in connection with FIG. 3, over the top
aperture 48 of the slab 46, chopped (e.g. at a frequency of 3.03
kHz) and quasi-collimated radiation emanating from the LED 70 (as
driven by driver circuit 24) through the apertures 68 and 48 will
illuminate the bottom contour of the finger 78. The use of a
specifically chopped radiation is needed in order to suppress the
influence of stray background radiation from the surroundings and
greatly improves the signal-to-noise (S/N) performance of the
sensor unit 44. The reflected radiation will be detected by the
annular silicon photodiode 72 located beneath the aperture 48. The
amount of reflected chopped radiation received by the photodiode 72
in essence will map out the topographical contour of the
fingerprint with the ridges reflecting more light and the troughs
less. FIG. 7 shows a typical fingerprint ridge and trough pattern
of a fingerprint along a typical scanned direction 1,2 (for
example) with the concomitant linear signature produced in (analog
form) developed by the sensor unit 44 and processed by circuit 24
shown in FIG. 8.
[0071] In addition to providing the appropriate driving pulses of a
particular designed frequency to the LED 70, the circuit 24 of the
SPA 20 (of FIG. 4) digitizes, after filtering and amplification,
the received analog signal waveform from the photodiode 72. This
amplitude-digitized waveform is then passed on to the
microprocessor 16 for further processing by a novel software
algorithm (to be described below) stored in same. A so-called
similarity index (SI) is generated from the processed candidate and
reference linear signatures which is then used to simply and
accurately authenticate the individual whose reference linear
signatures are submitted and stored earlier in the SPA 20.
[0072] Assuming that an individual has a priori his four linear
signatures taken and his Personal Code duly selected and provided
to proper authorities, as for example his employer, then in
operation of having his identity subsequently authenticated, he has
to first provide his identity (name and ID number) to the SPA
device via the use of a magnetic card using slot 30 or a smart
card. He then waits for the ready signal from the SPA unit 20
through the latter's LCD indicator 34 to scan his finger according
to his Personal Code. The rest of the operation is self-contained
and carried out by the SPA 20 without further intervention by the
individual. A successful authentication of the inndividual will
first br indicated by the LCD indicator 34 followed by appropriate
functionms to be performed by the security sentinel unit such as
the unlatching of a lock, the opening of a gate, etc.
[0073] The novel software algorithm developed for the current
invention is best described using an alternate embodiment of the
present invention where the two-step process described above is
combined into just one step. In other words, both the reference and
candidate linear signatures referred to earlier of the individual
to be authenticated will be procured with the same or functionally
similar hardware device, namely in this present case, by the
Selective Partial-fingerprint Authenticator (SPA). In this
preferred embodiment, an individual's "reference" Personal Choice
Biometric Signature (PCBS), for example two linear signatures with
a particular sequence as discussed earlier, will first be obtained
via an SPA device at an appropriate location such as a credit card
issuing bank. During a subsequent authentication of the individual,
e.g. at a point-of-sale terminal, ATM or other authentication
venues, a functionally similar SPA device having one's "reference"
PCBS stored a priori will be used to obtain one's "candidate" PCBS
for comparison and subsequent authentication. The advantage of
combining the two-step process into just one step is the fact that
since the same type of SPA devices are used to procure both the
"reference" and "candidate" PCBS's, the spatial variations of one's
linear signatures will be kept to a minimum.
[0074] FIG. 9A is taken from an actual computer printout and shows
a typical digitized linear signature of an individual obtained by
using an SPA device for a WEST to EAST scanning direction, namely
1,2 or {1} (see FIG. 3). The ordinate "LS.sub.k" denotes the linear
signature reflected signal from the finger (underside) in arbitrary
amplitude units as received by the silicon photodiode detector, and
the index "k" in the abscissa and the subscript "k" in the ordinate
denote digitized channel number ranging from "0" to "307" in FIG.
9A. In other words, the entire analog signal as a function of time
received from the detector during the scanning of the finger is
digitized into 308 channels each representing a fixed time period,
e.g. in the present case, one millisecond. For the data shown in
FIG. 9A, the SPA device has an aperture opening of 100 microns or
0.004".
[0075] Because of the typical finger curvature at both edges of the
finger, the data obtained from these edge regions are generally
less reliable due to the finger's incomplete contact with the
template, so a portion of the data is best deleted from the
beginning and end of the digitized linear signature shown in FIG.
9A. When this is done, as shown in dashed lines in FIG. 9A and
indicated by the points "a" and "b" of FIGS. 9A and 9B, the number
of channels is now reduced from 308 to 256 as illustrated in FIG.
9B. FIG. 9B now represents a typical digitized linear signature to
be obtained from an individual as both a reference and a candidate
PCBS in the current invention. The authentication process for the
individual involves the determination of whether an individual's
reference PCBS is the same as an individual's candidate PCBS.
Alternatively stated, if one's reference PCBS is found to be
sufficiently similar to that of one's candidate PCBS, then the
individual is considered authenticated.
[0076] Referring now to FIG. 10, a novel methodology is advanced in
the present invention in order to determine simply and reliably the
similarity between two digitized linear signatures, as typically
represented by FIG. 9B, for the purpose of authenticating an
individual. The formulation of a Similarity Index SI is advanced
following the procedural steps described by a flow chart 100. The
method starts out by preparing two digitized linear signatures
LS.sub.k(1) and Ls.sub.k(2), (described and shown in FIG. 9B above)
which are to analyzed or compared for similarity. Step 1 of the
formulation procedure for SI is to calculate the Fast Fourier
Transform FFT.sub.k1 and FFT.sub.k2, respectively, for both
LS.sub.k(1) and LS.sub.k(2) which are expressed and treated as
column vectors having "k" rows and one element per row, again as
shown in FIG. 9B. Note that both FFT.sub.k1 and FFT.sub.k2 are
complex vectors.
[0077] Step 2 of the formulation procedure is to delete the zero
frequency component from both of FFT.sub.k1 and FFT.sub.k2
obtaining, respectively, the vectors FFT.sub.k1-0 and FFT.sub.k2-0.
Step 3 of the formulation procedure is to calculate the normalized
Fast Fourier Transforms NFFT.sub.k1-0 and NFFT.sub.k2-0. This is
done by dividing, respectively, FFT.sub.k1-0 and FFT.sub.k2-0 each
by the square root of the dot product of itself (a complex vector)
with its own complex conjugate. As an example, FIG. 9C shows the
real part of a normalized complex vector NFFT.sub.k or FF.sub.k of
a digitized linear signature LS.sub.k as shown in FIG. 9B plotted
as the ordinate versus the digitized channel number "k" as the
abscissa. Note that the channel number "k" plotted for FF.sub.k in
FIG. 9C, extends only to 128 as the values for higher channel
number above 60 or so are all very close to zero.
[0078] The process of actually comparing the similarity or the two
original digitized linear signatures LS.sub.k(1) and LS.sub.k(2)
starts in step 4 where the dot product of NFFT.sub.k1-O and the
complex conjugate of NFFT.sub.k2-0 is calculated to be FFT.sub.k12
which is also a complex vector It can readily be shown that
FFT.sub.k12=FFT.sub.k21 which is the dot product of NFFT.sub.k2-0
and the complex conjugate of NFFT.sub.k1-0. Finally in step 5 one
formulates the Similarity Index SI as the dot product of
FFT.sub.k12 with its complex conjugate. Alternatively, SI can also
be formulated as and is equal to the dot product of FFTF.sub.k21
and its complex conjugate. Thus SI is deliberately formulated as a
real number ranging theoretically from "0" to "1". However, since
only the so-called Discrete Fourier Transform methodology is used
to calculate the FFT.sub.k's from the digitized linear signatures
LS.sub.k's (i.e., the integrating time is not extended to "+" and
"-" infinity), the value of SI for a perfect match or similarity
only approaches unity as a limit. Similarly, the value of SI for a
perfect mismatch or dissimilarity only approaches zero as a limit
Notwithstanding, this novel Similarity Index SI can be used as a
simply way to quantitatively ascertain the extent of similarity or
dissimilarity between two linear signatures through their
respective Fast Fourier Transforms and their subsequent formulation
procedures to obtain SI. Thus if the derived value of SI lies
closer to unity, the two linear signatures have more similarity and
vice versa.
[0079] As a working convenience, the "0" to "1 " range of SI is
divided into three bands to facilitate quantifying the output
results of the FIG. 10 formulation steps. An upper band including
SI results in the range of 0.7 to .about.1 is considered to be a
positive authentication outcome, leading to a verification
declaration, thereby confirming the identity of the individual in
question. A lower band including SI results in the range of
.about.0 to 0.3 is considered to be a negative authentication
outcome, leading to a lack of verification, or a non-authentication
declaration for the individual whose linear signatures are
compared. A central band (SI>0.3 and<0.7) is considered an
indeterminate result, which has been found to be extremely unlikely
due to the efficacy of the FFT/DFT method employed. Clearly, the
upper and lower bands may be defined by a pair of predetermined and
preset threshold values (an upper and a lower) and these values may
be adjusted to increase the likelihood of a desired, well defined
outcome commensurate with the actual implementation of the SPA
devices employed.
[0080] In using the FFT representation presented above, the
features or characteristics in real time of the linear signature
LS.sub.k are reflected principally in the low frequency components
of the FFT.sub.k. The high frequencies of the FFT.sub.k reflect
only the noise content of the linear signature LS.sub.k. As used
throughout, "k" here refers to the digitized channel number for
LS.sub.k and FFT.sub.k For the sake of simplicity in the above
formulation of the Similarity Index SI, we have only applied a
uniformly weighted factor to both FFT.sub.k1 and FFT.sub.k2 without
taking into consideration the effects of the weighted frequency
contribution to content and noise. A better formulation for the SI
will be to multiply both FFT.sub.k1 and FFT.sub.k2 with a
non-uniform weighted factor, such as an exponentially decaying
function, before forming the complex vector FFT.sub.k12 or
FFT.sub.k21. Such a formulation for SI would yield an even more
accurate similarity criterion for comparing linear signatures.
[0081] FIGS. 11A and 11B show, respectively, the reverse linear
signature LS.sub.k (i.e. scanning in the opposite direction) and
its FF.sub.k (or NFFT.sub.k) versus the digitized channel number
"k" for that particular linear signature LS.sub.k similar to that
shown in FIG. 9B. In other words, the linear signature shown in
FIG. 11A is obtained with the SPA device when the finger is
scanning in the EAST to WEST direction, 2,1 or {2} (see FIG. 3).
Even though the linear signatures LS.sub.k's for the two directions
appear similar, the value of SI when FIG. 9C is compared with FIG.
11B is calculated to be 0.0463 using the procedural steps of FIG.
10. Using the SI criterion as a measure of similarity match for
linear signatures LS.sub.k's, the two are very dissimilar because
the value is very much closer to zero, and clearly within the lower
band as defined above.
[0082] FIGS. 12A and 12B show, respectively, a slightly modified
linear signature LS.sub.k and its FF.sub.k (or NFFT.sub.k) versus
"k" for the particular linear signature LS.sub.k shown in FIG. 9B.
In this case, we purposely vary the amplitudes of some of the peaks
and valleys of the original linear signature LS.sub.k in FIG. 9B
representing some signal variations that might happen in a real
life situation even though the same procedure is used to obtain the
data with a functionally similar SPA device. The SI value comparing
the two linear signatures LS.sub.k's (FIG. 9B and FIG. 12A) is
calculated to be 0.9861. Thus, based upon the SI criterion for
matching, the two linear signatures are very similar because the
value is much closer to unity, as indeed expected, and clearly
within the upper band as defined above.
[0083] FIGS. 13A and 13B show, respectively, a new and dissimilar
linear signature LS.sub.k and its FF.sub.k (or NFFT.sub.k) versus
"k". As expected, the SI value between these two linear signatures
LS.sub.k (see FIGS. 13A and 9B) is calculated to be 0.0373. Thus
one can conclude from the SI criterion that these two linear
signatures LS.sub.k are very dissimilar, as expected.
[0084] FIGS. 14A through 14D show the effect of the loss of certain
amounts of data from the linear signatures LS.sub.k on the value of
the subsequently calculated Similarity Index SI. FIG. 14A
represents a reference linear signature with the data complete and
undeleted. FIGS. 14B, 14C and 14D represent, respectively, the loss
of over 15% of the contiguous data from FIG. 13A in the beginning,
in the end and symmetrically from the beginning and the end. Note
that in all four cases the scale of the time axis remains unchanged
indicative of the fact that the finger movement speed through the
sensor of the SPA unit is maintained relatively constant. The SI
values for the three cases when compared with the reference linear
signature shown in FIG. 14A are calculated respectively to be
0.8257, 0.8815 and 0.8637. Thus from the SI criterion of similarity
matching, the effect of loss of data from the reference or
candidate linear signatures is indeed small as long as the finger
movement speed across the measuring sensor is maintained relatively
constant.
[0085] Indeed linear signatures of the same finger measured by the
same SPA unit but with different finger movement speeds (up to a
factor of 10) can exhibit widely different calculated SI values. In
general, the quality of the SI value used to ascertain similarity
worsens as the finger movement speed deviates more and more from
the nominal speed. In order to overcome this problem, the sensor in
the SPA unit is always gated to sample the reflected light from the
bottom of the finger at a constant interval or period, e.g. once
every millisecond. Assuming now that for a nominal finger movement
speed, a total of M data points are generated for each pass of the
finger through the sensor that measures the reflected light. If the
finger movement speed is faster than the nominal speed, the number
of data points acquired for each pass will be smaller than M. If on
the other hand, the finger movement speed is slower, the number of
data points acquired will be greater than M.
[0086] The design of the SPA device unit specifies the range of
acceptable finger movement speed so as to limit the maximum and
minimum number of data points to be collected. In other words, the
SPA unit will show an invalid signal light when the finger movement
speed lies outside of this limit during data taking. If such a case
occurs, the individual would have to re-pass his finger through the
sensor until the SPA accepts the data to be collected. It is then
up to the resident software algorithm to automatically scale the
time axis for the reference and candidate linear signatures. If the
measured candidate linear signature has less than M data points
(finger movement too fast as compared with the nominal speed), the
software algorithm will adjust to reduce the number of reference
linear signature data points in order to achieve the same time axis
scale for matching comparison with the candidate linear signature
using the SI scheme. Similarly, if the measured candidate linear
signature has more than M data points (finger movement too slow as
compared with the nominal speed), the software algorithm will
adjust to reduce the number of candidate linear signature data
points in order to maintain the same time axis scale.
[0087] Referring now to FIGS. 15A through 15D, we consider now the
effect of the aperture size of the SPA device on the quality of the
SI similarity criterion. By doubling the aperture size from 100
microns (0.004") to 200 microns (0.008"), the measured linear
signatures LS.sub.k versus "k" as shown in FIG. 9B and FIG. 13A
become those shown in FIGS. 15A and 15C. Comparing these linear
signatures LS.sub.k procured with a larger aperture (FIGS. 15A and
15C) with the same linear signatures LS.sub.k procured with a
smaller aperture (FIG. 9B and FIG. 13A), one can see that although
the overall signal has increased, the contrast (i.e. the height
between the peaks and valleys) has decreased, which is as expected
due to a larger integrating effect over the reflected surface
experienced by the detector when the aperture becomes larger. The
value of SI for matching the linear signatures LS.sub.k in FIG. 9B
and FIG. 13A (smaller aperture) versus those in FIGS. 15A and 15C
(larger aperture) shows only a slight difference, namely from
0.0260 to 0.0210, and still shows a great dissimilarity between the
two respective linear signatures, leading to a lack of
authentication declaration.
[0088] FIGS. 15B and 15D are the reverse linear signatures (i.e.
scanning in the opposite direction) of those shown in FIGS. 15A and
15C, all procured with a larger aperture in the SPA device. The
calculated SI value for the linear signatures shown in FIGS. 15A
and 15B is 0.0300. This compares with the value of SI for the same
linear signatures but with a smaller aperture of 0.0460 and still
showing great dissimilarity as expected. The calculated SI value
for the linear signatures shown in FIGS. 15C and 15D is 0.112. This
compares with the value of SI for the same linear signatures but
with a smaller aperture of 0.0490 and still showing great
dissimilarity.
[0089] Referring now to FIGS. 16A through 16C, note, as expected,
that when finger movement is slower that the nominal speed, there
will be more data points collected since each data point is always
allotted a fixed time period, e.g. 1 millisecond. In the case shown
in FIG. 16B, since the speed is too slow (e.g. half the nominal
speed), there will be twice as many sample data points (260)
collected as in FIG. 16A (130). Similarly in the case for FIG. 16C,
since the speed is too fast (e.g. twice the nominal speed), there
will be only half as many sample data points collected (65) as in
FIG. 16A (130).
[0090] In order to invoke the Similarity Index methodology to
compare the similarity of the too slow candidate linear signature
shown in FIG. 16B with the nominally correct speed reference linear
signature shown in FIG. 16A, one has to adjust the time scale of
the candidate linear signature to be the same as that of the
reference linear signature. In this case, this may be achieved by
retaining one out of every two data points for the candidate linear
signature shown in FIG. 16B. After this is done, the value of SI
between the candidate linear signature taken at half speed and the
reference linear signature taken is calculated using the
formulation steps of FIG. 10 to be>0.95, indicating that the two
linear signatures are very similar, and authentication may be
declared.
[0091] To do the same for the too fast candidate linear signature
shown in FIG. 16C, one may adjust the time scale of the the
reference linear signature by taking out one of every two data
points from the reference linear signature shown in FIG. 16A. After
this is done, the value of SI between the candidate linear
signature taken at twice the speed and the reference linear
signature is calculated by the method of FIG. 10 to also
be>0.95, again indicating that the two signatures are very
similar, from which authentication can be declared.
[0092] Thus one can see that by adjusting the time scales for the
reference and candidate linear signatures, the Similarity Index
method of comparing similarity is rendered independent of the
finger movement speed, even for the continuum of finger speeds
normally encountered, as long as the number of sample data points
are within the above described acceptable finger movement speed
range. At this point it is worth noting that the several plots of
linear signature amplitudes vs. "k" and their corresponding Fourier
spectra plots resulted from actual computer printouts. During
system evaluations, these several plots were carried out using a
range of system parameters and the "k" indexes are therefore not
necessarily drawn to exactly the same baseline system parameters.
Thus, the number of data points of FIGS. 16A, 16B and 16C--130,
260, and 65--may not necessarily coincide with the number of data
points of FIGS. 9A, 9B, or any of the other plots, and the "too
fast" or "too slow" explanations in connection with FIGS. 16A-16C
do not apply to or impact on the number or significance of the data
points shown in the other plots.
[0093] From the discussion above, one can readily conclude that the
Similarity Index SI advanced in the current invention and described
earlier represents an excellent similarity matching indicator for
linear signatures. This SI is relatively independent of small
changes of peaks and valleys amplitudes in the measured linear
signatures. It is also relatively independent of any loss of data
in the acquired candidate linear signatures. By designing the
corresponding Selective Partial-fingerprint Authenticator SPA with
a fixed data sampling period, the finger movement speed across the
sensor can be rendered completely independent with the aid of a
simple scaling algorithm resident in the SPA. Furthermore, the time
it takes to acquire either one's reference or one's candidate PCBS
is very small (typically less than a second) and the time required
to authenticate two PCBS's is also very small (.about.1 second)
thus rendering this authentication technique very fast indeed.
Since the SPA device unit used in conjunction with this PCBS
methodology employs only one single LED source and one sensor
element, it is very low-cost when compared with other complete
fingerprint authentication methods which require the use of either
a line or a 2-dimensional sensor array for capturing the complete
candidate fingerprint.
[0094] Finally, as far as the level of security attainable using
the current PCBS authentication methodology is concerned, it is a
function of the number of linear signatures used (the number N
defined earlier) and the probability that two individuals will have
the same linear signature in any one of the scanning direction,
namely {1}, {2}, {3}or {4} (see FIG. 3). Assuming that one out of
100 individuals has the same linear signature when scanned in a
particular direction (e.g. for a value of SI=0.7 or higher), the
security level reached with the use of two sequential linear
signatures is already 1 in 10 to the 4th, the same as the PIN
methodology widely in use today. However, the current SI
methodology has the added security in the personal choice of 1 in
16 (assuming N=2) in picking any two sequential linear signatures
for one's own authentication, thus the level of security reachable
by the current invention is 16 times more secure than the PIN
counterpart If one cares to pick a higher number for N, e.g. N=4,
then the security level is a factor of 2.56.times.10 to the 6th,
more secure than the current PIN methodology.
[0095] In addition to its application as a security sentinel
described above, the current invention has numerous other
applications primarily in the security industry. These include
locks of all kinds such as door lock, automobiles locks, safes etc.
But one of the most promising applications is in the realization of
the so-called biometric smart card discussed earlier in the prior
art section of this patent application. In order to accomplish this
here-to-fore unrealized capability, all the components contained in
the SPA device as shown in FIG. 4, with the exception of the
magnetic reader head 28 and smart card receptacle 32, have to be
incorporated within the confines of a regular-sized credit
card.
[0096] This can be done by first incorporating the sensor
driver/signal pre-processor circuit 24 and the microprocessor 26 as
part of the smart chip which is already resident in the smart card.
The smart chip now takes over the functions of both the sensor
driver/signal pre-processor 24 and the microprocessor 26.
Alternatively, a MEMS-based ASIC chip combining all the components
of the sensor unit 44 could also include the functions of the
sensor driver/signal-processor circuit 24. The power supply circuit
36 and the battery unit 40 have to be replaced with the latest and
fast maturing super-thin Power Paper Cell battery which could be
easily incorporated within the size and thickness of an ordinary
credit card.
[0097] Finally the LCD 34 indicator's function is replaced with an
LED 70A (of FIG. 5) in the sensor unit 44 without altering its
original illumination capability. The LED 70 is converted from a
one-color into a "three-color" LED device 70A. The first color
(red) is used as a blinking light operating at a very slow
frequency like 0.33 Hz or once every three seconds. By a user
simply blocking this blinking red LED light momentarily, this LED
70A will turn into a steady yellow color indicative that the card
is ready to accept the linear signatures from the individual
according to his Personal Code. This can be done because as the
blinking red LED is blocked, the silicon photodiode will receive a
sizable signal due to the sudden presence of a large amount of
reflected light. If the submittal of the linear signatures by the
individual according to his own Personal Code is accepted by the
card, then the LED 70A will turn into a steady green color
indicative that the holder of the biometric smart card has been
authenticated and the smart card can now be accepted for financial
(credit, debit, etc.) and other transactions. If the linear
signatures are not accepted, the LED will turn back into a blinking
red color again prompting the cardholder to repeat the
authentication process. Three unsuccessful submittal of linear
signatures by a cardholder will make the LED steady red and the
biometric smart card will no longer be valid until it is reset by
the issuer or a proper authority.
[0098] Although the present invention has been described in terms
of preferred and several alternate embodiments, the invention
should not be deemed limited thereto since other embodiments and
modifications will readily occur to one skilled on the art. It is
therefore to be understood that the appended claims are intended to
cover all such modifications as fall within the true spirit and
scope of the invention.
* * * * *